[GSoC 2026]Project #8: Local Obstacle Avoidance for Dora Robots

[sqw420]

Hello,dora-rs team!
I am interested in the "Local Obstacle Avoidance for Dora Robots" GSoC project idea and wanted to share what I've built so far and get your feedback before finalizing my proposal.
Objective: To develop a local planner operator for the Dora framework that enables real-time obstacle avoidance and smooth navigation in a 2D simulated environment.
Core Logic Selection: Comparison between Dynamic Window Approach (DWA) and Artificial Potential Field (APF) for implementation.
Key Questions for Clarification:

● What is the specific format of the obstacle input data (raw sensor vs. processed clusters)?
● Are we prioritizing computational speed (APF) or trajectory optimality (DWA)?
● How is the local goal generated (global planner integration vs. static input)?
● What are the exact motion constraints of the simulated robot (differential vs. omnidirectional)?
● Is real-time graphical visualization of the trajectory required for the demo?

For this project, I have implemented a simple DWA-based local planner and integrated it into the dora_nav open-source project. The results are as follows:

Demo Video:https://www.bilibili.com/video/BV1z3XoBdEWA/?vd_source=5103ee22355f0209eaf90b2304ac2fb9

Looking forward to your feedback!
Thanks

[RuPingCen]

Thanks for reaching out and sharing your progress—your current work is really impressive, and it’s great to see you’ve already formed a clear design around the local obstacle avoidance task.
To clarify the context first: this local planner is a key component within the entire Dora robot navigation system. The global path planner will output a reference global path, and our goal is to build the local planner to perform real-time obstacle avoidance while tracking that global path. Given this requirement, we recommend prioritizing the DWA (Dynamic Window Approach) for its better trajectory smoothness, motion constraint compliance, and practical performance in mobile robot navigation.